Generating a taxonomy for documents from tag data
View Patent ↗A method and system for generating a taxonomy for documents from tag data are provided. The method includes obtaining tag data in the form of tags on documents with tag weightings for a document and clustering the tags using the tag weightings for documents, wherein each cluster is an identified subject. The documents are associated with each identified subject and the subjects are compared to identifying relationships between subjects to build a taxonomy of subjects. A tag weighting for a document is the number of times the tag is applied to the document with a user rating of the relevance of the tag to the document. The steps are carried out automatically without user intervention.
1. A method for generating a taxonomy for documents from tag data, comprising:
obtaining tag data in the form of tags on documents with tag weightings for a document;
clustering the tags using the tag weightings for documents, wherein each cluster is an identified subject;
associating the documents with each identified subject;
comparing the subjects and identifying relationships between subjects; and
building a taxonomy of subjects using the relationships between subjects;
wherein the obtaining, the clustering, the associating, the comparing, and the building are carried out automatically without user intervention.
2. The method as claimed in claim 1 , wherein building a taxonomy is dynamically updated to reflect a snapshot of tag data at a given time.
3. The method as claimed in claim 1 , wherein a tag weighting for a document is the number of times the tag is applied to the document.
4. The method as claimed in claim 1 , wherein tag weighting for a document includes user ratings of the relevance of the tag to the document.
5. The method as claimed in claim 1 , wherein clustering the tags includes cutting an obtained dendrogram to produce clusters.
6. The method as claimed in claim 1 , including:
identifying homonyms in the tags, and replacing homonym tags with different names to provide multiple versions of the same tag, one for each of the different contexts.
7. The method as claimed in claim 1 , including:
identifying synonyms in the tags.
8. The method as claimed in claim 1 , wherein associating the documents with each identified subject includes:
determining the key tags clustered in a subject; and
determining if the average weighting of the key tags in a document is above a predefined threshold.
9. The method as claimed in claim 1 , wherein associating the documents with each identified subject includes:
correlating tag vectors of a subject and a document.
10. The method as claimed in claim 1 , wherein comparing the subjects and identifying relationships between subjects, includes:
determining the extended tags of a subject as all the tags of documents tagged by a key tag in a subject;
comparing the extended tags of a pair of subjects.
11. The method as claimed in claim 1 , wherein comparing the subjects and identifying relationships between subjects, includes:
relating subjects to each other through the documents with which they are associated.
12. The method as claimed in claim 1 , wherein building a taxonomy of subjects, includes:
using weights of associations between subjects to determine the likelihood of a relationship.
13. The method as claimed in claim 1 , wherein building a taxonomy of subjects, includes:
providing the subjects as a network of nodes and converting the network into a tree.